Personality & Individual Differences · Simplified
Three Findings, Two Lines
Whether sensitivity is a vulnerability, an advantage or both can turn entirely on where somebody happened to look.
What is absent
What is absent here. No gene is named and no genotype appears: the grouping variable is a fictional sensitivity score. No disorder, diagnosis or clinical risk estimate is used, and the outcome axis is an unnamed standardised outcome. A slope on this page is an average tendency across a group and never a prediction about a person, and no figure here corresponds to any published finding.
Step 1 of 2
First, a judgement
A study measures two groups, one scoring higher and one lower on a sensitivity score, against the quality of their environment. Within the range it sampled, the higher-sensitivity group does worse, and the two groups converge towards the better end of what was measured. Which of these does that support?
Step 2 of 2
The same two lines, three different studies
The two lines below do not change at any point on this page. What changes is only which part of the environmental range a study happened to sample. The part inside the window is what that study saw; the faint part is what it did not.
Key terms
- Diathesis-stress
- Sensitivity as a vulnerability: worse outcomes in poor conditions, no advantage in good ones.
- Differential susceptibility
- Sensitivity working in both directions: worse in poor conditions and better in good ones. The two lines cross.
- Vantage sensitivity
- Sensitivity as an advantage: no difference in poor conditions, a larger benefit as conditions improve.
- Range restriction
- Having measured only part of a variable. It does not announce itself, and it is a property of the study rather than of the phenomenon.
Two lines, one per group, drawn across the whole environmental range. The solid stretch with points on it is what this study measured. The faint stretch is what it did not, and no study can see whether its own faint stretch matters.
What this shows
The theory you find is partly a fact about where you looked
Key idea: Three studies, three different published conclusions about what sensitivity is, and one underlying pair of straight lines that never changed. A study run in poor conditions finds a vulnerability. One run in good conditions finds an advantage. One that happens to span the point where the lines meet finds that sensitivity works both ways. Each is an accurate description of what it measured, and each becomes a claim about sensitivity in general only by an extrapolation that its own data cannot support. The uncomfortable part is that range restriction does not announce itself: nothing inside the first study's window looks wrong, the lines are straight, the interaction is real, and the statistics are correct. What tells you is knowing what range was sampled and refusing to read past its edges, which is why a paper reporting an interaction should report the range it observed as carefully as it reports the interaction.
Both lines here are perfectly straight and the data have no curvature, no measurement error to speak of and no confounding, which is what lets the three patterns separate so cleanly. Real data are far messier and the distinction between these accounts is genuinely hard to establish; the three are actively argued about, and showing that a restricted range can manufacture the appearance of one of them is not evidence that any particular published finding is an artefact. Two further cautions. A slope is an average tendency across a group and says nothing about what any individual in it will do, and the groups here are two halves of a fictional continuous score rather than two kinds of person. And the direction the argument runs matters: widening the range can reveal an interaction that was hidden as easily as it can dissolve one that looked clear.
The longer version lets you move the crossover point and both slopes independently, set the window's width as well as its position, and work through what each theory predicts before you see any data. It is at Gene by Environment Interaction Visualiser in the main collection.